Where the numbers come from

1. Every value is fetched, never invented

Panels read live upstream APIs. There is no demo dataset, no seeded sample content and no back-filled history in production. The publication records in Nebula are real curated papers with DOI links; the trials, approvals and quotes in PharmaTerminal are transcribed from the registries and feeds named on each panel.

2. Unavailable beats estimated

If an upstream source is down, the interface says the data is unavailable. It does not show the last known value as if it were current, and it does not substitute a plausible-looking placeholder. You will see this behaviour on this very site: where a feed is missing, the panel states that rather than rendering a number.

3. Caveats sit on the field they concern

Where a value has a meaningful limitation, the limitation is attached to that value — for example the label effective date on a drug-labelling record, or the unknown-value convention on a trial field — rather than being relegated to a general disclaimer at the bottom of the page.

4. Sources are credited, not implied as endorsements

openFDA, ClinicalTrials.gov, PubMed/NCBI, the prediction-market venue and the publisher feeds are named as sources. None of them endorses, sponsors or is affiliated with this suite, and nothing here should be read as their statement.

openFDA — Drugs@FDA

Used for: PDUFA / action dates and recent authorisations: submission status dates, sponsor, route and submission type.

Limits: Reflects what has been filed and recorded, not what will happen. An action date is a scheduled date, and schedules move. Nothing here predicts an outcome.

open.fda.gov/

ClinicalTrials.gov (API v2)

Used for: Registered studies: NCT identifier, phase, indication, enrolment, status and completion date.

Limits: Registry entries are sponsor-reported and vary in completeness and freshness. Fields genuinely absent at source are shown as unknown — the products display “Unknown” rather than inferring a value.

clinicaltrials.gov/

NCBI PubMed / E-utilities

Used for: Publication records behind Nebula's evidence maps and the publication layer: title, authors, journal, year, DOI and abstract.

Limits: A publication record is a record of publication, not of quality. Nebula additionally discloses that its edge set mixes deterministic keyword overlap with random sampling, and that citation data is partially sparse.

pubmed.ncbi.nlm.nih.gov/

Prediction markets (Polymarket)

Used for: The R&D prediction-markets panel: keyword-matched markets on therapeutic questions.

Limits: Only markets matched to pharma and R&D keywords are listed; unrelated markets are deliberately not substituted in to fill the panel. A market price is a crowd's current opinion, not a forecast we endorse.

polymarket.com/

Market data feed

Used for: Quotes, percentage changes, volumes and short trend lines for large-cap pharma and sector indices.

Limits: Delayed quotes. Shown to demonstrate monitoring, not to trade on. We publish no price predictions and no success probabilities.

Publisher news feeds

Used for: The breaking-pharma-news panel. Headlines and timestamps are reproduced as published, tagged by event type.

Limits: Inclusion reflects what a feed published, not editorial selection or verification by us. Where a source is unreachable the feed is reported as unavailable.